You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2021/12/23 09:47:30 UTC
[GitHub] [tvm] ashutosh-arm opened a new pull request #9801: [CMSIS-NN] Conv2D with equal paddings can be mapped to CMSIS-NN target
ashutosh-arm opened a new pull request #9801:
URL: https://github.com/apache/tvm/pull/9801
Bugfix for padding support for Conv2D via CMSIS-NN.
TFLu and CMSIS-NN support symmetric padding in Conv2D.
This commit modifies:
-pattern matching for Conv2D padding
-utility function that provides padding values to unit tests
-a negative test to test asymmetric padding does not work
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] Mousius merged pull request #9801: [CMSIS-NN] Conv2D with equal paddings can be mapped to CMSIS-NN target
Posted by GitBox <gi...@apache.org>.
Mousius merged pull request #9801:
URL: https://github.com/apache/tvm/pull/9801
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] ashutosh-arm commented on a change in pull request #9801: [CMSIS-NN] Conv2D with equal paddings can be mapped to CMSIS-NN target
Posted by GitBox <gi...@apache.org>.
ashutosh-arm commented on a change in pull request #9801:
URL: https://github.com/apache/tvm/pull/9801#discussion_r778197763
##########
File path: tests/python/contrib/test_cmsisnn/test_conv2d.py
##########
@@ -70,7 +70,7 @@ def make_model(
invar = relay.var("input", shape=shape, dtype=dtype)
p = (0, 0, 0, 0)
if padding == "INVALID":
- p = [1, 2, 1, 2]
+ p = [1, 2, 2, 1]
Review comment:
not exactly the same, but a similar one will capture the non-matching padding.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] Mousius commented on a change in pull request #9801: [CMSIS-NN] Conv2D with equal paddings can be mapped to CMSIS-NN target
Posted by GitBox <gi...@apache.org>.
Mousius commented on a change in pull request #9801:
URL: https://github.com/apache/tvm/pull/9801#discussion_r776256371
##########
File path: tests/python/contrib/test_cmsisnn/test_conv2d.py
##########
@@ -70,7 +70,7 @@ def make_model(
invar = relay.var("input", shape=shape, dtype=dtype)
p = (0, 0, 0, 0)
if padding == "INVALID":
- p = [1, 2, 1, 2]
+ p = [1, 2, 2, 1]
Review comment:
Given we've changed this case, do we have the previous behaviour captured in the non-matching test to ensure it's no longer matched?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] Mousius commented on pull request #9801: [CMSIS-NN] Conv2D with equal paddings can be mapped to CMSIS-NN target
Posted by GitBox <gi...@apache.org>.
Mousius commented on pull request #9801:
URL: https://github.com/apache/tvm/pull/9801#issuecomment-1002524922
Overall looks good, I think we're just missing one test case :smile_cat:
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org